dataset.py, which recomputes what the entry states: the script block by block · open in ColabDictionary of Applied Machine Learning: 115 machine-learning terms, each with a definition, a figure, citations and a Python demo · All terms · Demos · Start here
Updated on 2026-10-29
Readiness 100/100
See also data point data feature bootstrap sample size
A dataset is a finite collection of data points on which model training and validation are performed. Strictly speaking, a dataset is a set of distinct data points. With slight abuse of terminology, the term is also used interchangeably with sample, a finite sequence that may contain repetitions. Read as such a sequence, a dataset represents a probability distribution through its empirical distribution, in which a repeated data point counts as often as it is listed. The size of a dataset is measured most simply by its number of data points, which overstates what correlated data points carry and says nothing about how well they cover the feature space. In practice, machine learning (ML) methods do not access the data points themselves but a representation of them, such as a table of the relational model, which records their attributes and nothing about the dataset's fitness for a purpose.
dataset.py, which recomputes what the entry states: the script block by block · open in ColabP-setsampleThe cows of Fig. 2,
or the rows of Table 1 that record them, form a
dataset. A dataset is a collection
$\mathcal{D} = \{{\mathbf z}^{(1)}, \ldots, {\mathbf z}^{(m)}\}$ of
data points. Machine learning (ML) methods use a dataset for model
training and validation. Strictly speaking, a dataset is an
unordered collection of distinct data points, i.e., a set with no
repetitions. In ML literature, however, the term is often used for
a sample: a sequence of $m$ data points,
indexed by their position $r = 1, \ldots, m$ in
the sequence, that may contain
repetitions (see Fig. 1).
The size of a dataset has more than one measure. The most basic one is the number $m$ of data points in the list. When the data points are realizations of correlated random variables (RVs), however, $m$ overstates what they carry: the average over $30$ consecutive days of a weather record scatters as much as the average over $7$ days drawn at random from the whole record, because consecutive days are correlated (see sample size). The size of a dataset is then defined only relative to a probabilistic model of the data generation, as the number of independent and identically distributed (i.i.d.) data points that would carry the same information. A third measure asks how well the data points cover the feature space: the largest distance from a point of the region of interest to the nearest data point (see covering), which is small for a dataset whose features are spread out and large for one that leaves parts of the feature space empty, however many data points it holds.
The notion of a dataset is broad: data points may represent
concrete physical entities (such as humans or animals) or abstract
objects (such as numbers). For illustration,
Fig. 2 shows a dataset whose
data points are cows.
P-tableA single table consists of rows and columns, where each row corresponds
to a single data point and each column represents a specific
attribute of a data point. The order of rows does not matter, and each
attribute is associated with a domain that specifies its set of
allowed values. ML methods use these attributes as the
features or the label of a data point, so the attribute
domains correspond to the feature space and the
label space. Table 1 shows an example obtained
from the dataset in Fig. 2.
Trustworthy artificial intelligence (trustworthy AI) also requires documenting how a dataset was assembled, for example through a datasheet that records the dataset's motivation, composition, and collection process (Gebru et al., 2021). When the data points are personal data, this becomes a legal duty: under the purpose limitation principle, the general data protection regulation (GDPR) requires that such data be collected for "specified, explicit and legitimate purposes" (European Parliament and Council of the European Union, 2016, Art. 5(1)(b)). These purposes are fixed at the time of collection.
See also: data point, sample, data, relational model, feature, empirical distribution, bootstrap, sample size, covering, EU AI Act.
Published version: doi.org/10.1007/978-981-92-2012-0_10-1, in Dictionary of Applied Machine Learning (Springer Nature Singapore). Please cite that version. This page corresponds to version 1 of that entry.
@incollection{dictml_dataset,
author = {Jung, Alexander},
editor = {Olioumtsevits, Konstantina and Schnoor, Ekkehard},
title = {dataset},
booktitle = {Dictionary of Applied Machine Learning},
publisher = {Springer Nature Singapore},
year = {2026},
doi = {10.1007/978-981-92-2012-0_10-1},
isbn = {978-981-92-2012-0},
url = {https://doi.org/10.1007/978-981-92-2012-0_10-1}
}